Virtual Scientific Communities for Driving Innovation

Presenter: Karl Lieberherr, Northeastern University, College of Computer and Information Science, Boston, MA.

Abstract:

Note to AOSD 10 Program Committee: This game proposes an interesting software development process for problem solving software. The researcher role is played by a software agent. The game works well for developing software for optimization problems. The developers improve each others software through competition/collaboration.

We present the Specker Challenge Game (SCG) and our experience in using it for driving innovation in a specific domain. Adopting an SCG-centric research process has the following benefits:
(1) it focuses researchers on a specific domain by defining a language for expressing hypotheses about that domain. Thus, reducing the amount of management effort.
(2) it provides a structured framework for collaboration between researchers. The researchers provide and receive frequent feedback on their hypotheses from their peers. This makes collaboration more effective.
(3) it accumulates knowledge in domain X. Which would be otherwise scattered in emails, programs and minds.
(4) researchers are motivated towards proposing and opposing non-trivial hypotheses in order to gain reputation. They like to win and if they lose, they want to find out why.
(5) managers get a fair comparison of the skills of their researchers through the competition results.
(6) controlled teaching and learning through the game. Researchers who introduce new knowledge entice other researchers to assimilate the same or better knowledge. Researchers that don't participate in this activity, lose reputation, as they would in a real scientific community.

To use SCG in a particular domain, a suitable hypothesis language needs to be designed. We will demonstrate two hypothesis languages for the Boolean Maximum Constraint Satisfaction Problem (Boolean MAX-CSP). We will also demonstrate how to play the game by proposing and opposing hypotheses.

Currently, SCG games are administrated via a web-based game administrator that we use effectively for teaching our software development course where the students implement Boolean MAX-CSP solvers. The students write programs, called agents, that form a virtual scientific community living on the web, controlled by the SCG rules. The winning agent, after a full round-robin tournament, has the most innovative Boolean MAX-CSP solver for this virtual community.

More information on SCG is available: SCG Home Page.

Joint work with Ahmed Abdelmeged and Bryan Chadwick